[deliverable/linux.git] / drivers / gpu / drm / amd / amdkfd / kfd_device.c

/*
 * Copyright 2014 Advanced Micro Devices, Inc.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
 * OTHER DEALINGS IN THE SOFTWARE.
 */

#include <linux/amd-iommu.h>
#include <linux/bsearch.h>
#include <linux/pci.h>
#include <linux/slab.h>
#include "kfd_priv.h"
#include "kfd_device_queue_manager.h"

#define MQD_SIZE_ALIGNED 768

static const struct kfd_device_info kaveri_device_info = {
	.max_pasid_bits = 16,
	.mqd_size_aligned = MQD_SIZE_ALIGNED
};

struct kfd_deviceid {
	unsigned short did;
	const struct kfd_device_info *device_info;
};

/* Please keep this sorted by increasing device id. */
static const struct kfd_deviceid supported_devices[] = {
	{ 0x1304, &kaveri_device_info },	/* Kaveri */
	{ 0x1305, &kaveri_device_info },	/* Kaveri */
	{ 0x1306, &kaveri_device_info },	/* Kaveri */
	{ 0x1307, &kaveri_device_info },	/* Kaveri */
	{ 0x1309, &kaveri_device_info },	/* Kaveri */
	{ 0x130A, &kaveri_device_info },	/* Kaveri */
	{ 0x130B, &kaveri_device_info },	/* Kaveri */
	{ 0x130C, &kaveri_device_info },	/* Kaveri */
	{ 0x130D, &kaveri_device_info },	/* Kaveri */
	{ 0x130E, &kaveri_device_info },	/* Kaveri */
	{ 0x130F, &kaveri_device_info },	/* Kaveri */
	{ 0x1310, &kaveri_device_info },	/* Kaveri */
	{ 0x1311, &kaveri_device_info },	/* Kaveri */
	{ 0x1312, &kaveri_device_info },	/* Kaveri */
	{ 0x1313, &kaveri_device_info },	/* Kaveri */
	{ 0x1315, &kaveri_device_info },	/* Kaveri */
	{ 0x1316, &kaveri_device_info },	/* Kaveri */
	{ 0x1317, &kaveri_device_info },	/* Kaveri */
	{ 0x1318, &kaveri_device_info },	/* Kaveri */
	{ 0x131B, &kaveri_device_info },	/* Kaveri */
	{ 0x131C, &kaveri_device_info },	/* Kaveri */
	{ 0x131D, &kaveri_device_info },	/* Kaveri */
};

static const struct kfd_device_info *lookup_device_info(unsigned short did)
{
	size_t i;

	for (i = 0; i < ARRAY_SIZE(supported_devices); i++) {
		if (supported_devices[i].did == did) {
			BUG_ON(supported_devices[i].device_info == NULL);
			return supported_devices[i].device_info;
		}
	}

	return NULL;
}

struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, struct pci_dev *pdev)
{
	struct kfd_dev *kfd;

	const struct kfd_device_info *device_info =
					lookup_device_info(pdev->device);

	if (!device_info)
		return NULL;

	kfd = kzalloc(sizeof(*kfd), GFP_KERNEL);
	if (!kfd)
		return NULL;

	kfd->kgd = kgd;
	kfd->device_info = device_info;
	kfd->pdev = pdev;
	kfd->init_complete = false;

	return kfd;
}

static bool device_iommu_pasid_init(struct kfd_dev *kfd)
{
	const u32 required_iommu_flags = AMD_IOMMU_DEVICE_FLAG_ATS_SUP |
					AMD_IOMMU_DEVICE_FLAG_PRI_SUP |
					AMD_IOMMU_DEVICE_FLAG_PASID_SUP;

	struct amd_iommu_device_info iommu_info;
	unsigned int pasid_limit;
	int err;

	err = amd_iommu_device_info(kfd->pdev, &iommu_info);
	if (err < 0) {
		dev_err(kfd_device,
			"error getting iommu info. is the iommu enabled?\n");
		return false;
	}

	if ((iommu_info.flags & required_iommu_flags) != required_iommu_flags) {
		dev_err(kfd_device, "error required iommu flags ats(%i), pri(%i), pasid(%i)\n",
		       (iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_ATS_SUP) != 0,
		       (iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_PRI_SUP) != 0,
		       (iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_PASID_SUP) != 0);
		return false;
	}

	pasid_limit = min_t(unsigned int,
			(unsigned int)1 << kfd->device_info->max_pasid_bits,
			iommu_info.max_pasids);
	/*
	 * last pasid is used for kernel queues doorbells
	 * in the future the last pasid might be used for a kernel thread.
	 */
	pasid_limit = min_t(unsigned int,
				pasid_limit,
				kfd->doorbell_process_limit - 1);

	err = amd_iommu_init_device(kfd->pdev, pasid_limit);
	if (err < 0) {
		dev_err(kfd_device, "error initializing iommu device\n");
		return false;
	}

	if (!kfd_set_pasid_limit(pasid_limit)) {
		dev_err(kfd_device, "error setting pasid limit\n");
		amd_iommu_free_device(kfd->pdev);
		return false;
	}

	return true;
}

static void iommu_pasid_shutdown_callback(struct pci_dev *pdev, int pasid)
{
	struct kfd_dev *dev = kfd_device_by_pci_dev(pdev);

	if (dev)
		kfd_unbind_process_from_device(dev, pasid);
}

bool kgd2kfd_device_init(struct kfd_dev *kfd,
			 const struct kgd2kfd_shared_resources *gpu_resources)
{
	unsigned int size;

	kfd->shared_resources = *gpu_resources;

	/* calculate max size of mqds needed for queues */
	size = max_num_of_processes *
		max_num_of_queues_per_process *
		kfd->device_info->mqd_size_aligned;

	/* add another 512KB for all other allocations on gart */
	size += 512 * 1024;

	if (kfd2kgd->init_sa_manager(kfd->kgd, size)) {
		dev_err(kfd_device,
			"Error initializing sa manager for device (%x:%x)\n",
			kfd->pdev->vendor, kfd->pdev->device);
		goto out;
	}

	kfd_doorbell_init(kfd);

	if (kfd_topology_add_device(kfd) != 0) {
		dev_err(kfd_device,
			"Error adding device (%x:%x) to topology\n",
			kfd->pdev->vendor, kfd->pdev->device);
		goto kfd_topology_add_device_error;
	}

	if (!device_iommu_pasid_init(kfd)) {
		dev_err(kfd_device,
			"Error initializing iommuv2 for device (%x:%x)\n",
			kfd->pdev->vendor, kfd->pdev->device);
		goto device_iommu_pasid_error;
	}
	amd_iommu_set_invalidate_ctx_cb(kfd->pdev,
						iommu_pasid_shutdown_callback);

	kfd->dqm = device_queue_manager_init(kfd);
	if (!kfd->dqm) {
		dev_err(kfd_device,
			"Error initializing queue manager for device (%x:%x)\n",
			kfd->pdev->vendor, kfd->pdev->device);
		goto device_queue_manager_error;
	}

	if (kfd->dqm->start(kfd->dqm) != 0) {
		dev_err(kfd_device,
			"Error starting queuen manager for device (%x:%x)\n",
			kfd->pdev->vendor, kfd->pdev->device);
		goto dqm_start_error;
	}

	kfd->init_complete = true;
	dev_info(kfd_device, "added device (%x:%x)\n", kfd->pdev->vendor,
		 kfd->pdev->device);

	pr_debug("kfd: Starting kfd with the following scheduling policy %d\n",
		sched_policy);

	goto out;

dqm_start_error:
	device_queue_manager_uninit(kfd->dqm);
device_queue_manager_error:
	amd_iommu_free_device(kfd->pdev);
device_iommu_pasid_error:
	kfd_topology_remove_device(kfd);
kfd_topology_add_device_error:
	kfd2kgd->fini_sa_manager(kfd->kgd);
	dev_err(kfd_device,
		"device (%x:%x) NOT added due to errors\n",
		kfd->pdev->vendor, kfd->pdev->device);
out:
	return kfd->init_complete;
}

void kgd2kfd_device_exit(struct kfd_dev *kfd)
{
	if (kfd->init_complete) {
		device_queue_manager_uninit(kfd->dqm);
		amd_iommu_free_device(kfd->pdev);
		kfd_topology_remove_device(kfd);
	}

	kfree(kfd);
}

void kgd2kfd_suspend(struct kfd_dev *kfd)
{
	BUG_ON(kfd == NULL);

	if (kfd->init_complete) {
		kfd->dqm->stop(kfd->dqm);
		amd_iommu_free_device(kfd->pdev);
	}
}

int kgd2kfd_resume(struct kfd_dev *kfd)
{
	unsigned int pasid_limit;
	int err;

	BUG_ON(kfd == NULL);

	pasid_limit = kfd_get_pasid_limit();

	if (kfd->init_complete) {
		err = amd_iommu_init_device(kfd->pdev, pasid_limit);
		if (err < 0)
			return -ENXIO;
		amd_iommu_set_invalidate_ctx_cb(kfd->pdev,
						iommu_pasid_shutdown_callback);
		kfd->dqm->start(kfd->dqm);
	}

	return 0;
}

void kgd2kfd_interrupt(struct kfd_dev *dev, const void *ih_ring_entry)
{
}
Commit	Line	Data
4a488a7a OG	1	/*
	2	* Copyright 2014 Advanced Micro Devices, Inc.
	3	*
	4	* Permission is hereby granted, free of charge, to any person obtaining a
	5	* copy of this software and associated documentation files (the "Software"),
	6	* to deal in the Software without restriction, including without limitation
	7	* the rights to use, copy, modify, merge, publish, distribute, sublicense,
	8	* and/or sell copies of the Software, and to permit persons to whom the
	9	* Software is furnished to do so, subject to the following conditions:
	10	*
	11	* The above copyright notice and this permission notice shall be included in
	12	* all copies or substantial portions of the Software.
	13	*
	14	* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
	15	* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
	16	* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
	17	* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
	18	* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
	19	* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
	20	* OTHER DEALINGS IN THE SOFTWARE.
	21	*/
	22
	23	#include <linux/amd-iommu.h>
	24	#include <linux/bsearch.h>
	25	#include <linux/pci.h>
	26	#include <linux/slab.h>
	27	#include "kfd_priv.h"
64c7f8cf	28	#include "kfd_device_queue_manager.h"
4a488a7a	29
19f6d2a6 OG	30	#define MQD_SIZE_ALIGNED 768
19f6d2a6 OG	31
4a488a7a OG	32	static const struct kfd_device_info kaveri_device_info = {
4a488a7a OG	33	.max_pasid_bits = 16,
19f6d2a6	34	.mqd_size_aligned = MQD_SIZE_ALIGNED
4a488a7a OG	35	};
	36
	37	struct kfd_deviceid {
	38	unsigned short did;
	39	const struct kfd_device_info *device_info;
	40	};
	41
	42	/* Please keep this sorted by increasing device id. */
	43	static const struct kfd_deviceid supported_devices[] = {
	44	{ 0x1304, &kaveri_device_info }, /* Kaveri */
	45	{ 0x1305, &kaveri_device_info }, /* Kaveri */
	46	{ 0x1306, &kaveri_device_info }, /* Kaveri */
	47	{ 0x1307, &kaveri_device_info }, /* Kaveri */
	48	{ 0x1309, &kaveri_device_info }, /* Kaveri */
	49	{ 0x130A, &kaveri_device_info }, /* Kaveri */
	50	{ 0x130B, &kaveri_device_info }, /* Kaveri */
	51	{ 0x130C, &kaveri_device_info }, /* Kaveri */
	52	{ 0x130D, &kaveri_device_info }, /* Kaveri */
	53	{ 0x130E, &kaveri_device_info }, /* Kaveri */
	54	{ 0x130F, &kaveri_device_info }, /* Kaveri */
	55	{ 0x1310, &kaveri_device_info }, /* Kaveri */
	56	{ 0x1311, &kaveri_device_info }, /* Kaveri */
	57	{ 0x1312, &kaveri_device_info }, /* Kaveri */
	58	{ 0x1313, &kaveri_device_info }, /* Kaveri */
	59	{ 0x1315, &kaveri_device_info }, /* Kaveri */
	60	{ 0x1316, &kaveri_device_info }, /* Kaveri */
	61	{ 0x1317, &kaveri_device_info }, /* Kaveri */
	62	{ 0x1318, &kaveri_device_info }, /* Kaveri */
	63	{ 0x131B, &kaveri_device_info }, /* Kaveri */
	64	{ 0x131C, &kaveri_device_info }, /* Kaveri */
	65	{ 0x131D, &kaveri_device_info }, /* Kaveri */
	66	};
	67
	68	static const struct kfd_device_info *lookup_device_info(unsigned short did)
	69	{
	70	size_t i;
	71
	72	for (i = 0; i < ARRAY_SIZE(supported_devices); i++) {
	73	if (supported_devices[i].did == did) {
	74	BUG_ON(supported_devices[i].device_info == NULL);
	75	return supported_devices[i].device_info;
	76	}
	77	}
	78
	79	return NULL;
	80	}
	81
	82	struct kfd_dev kgd2kfd_probe(struct kgd_dev kgd, struct pci_dev *pdev)
	83	{
	84	struct kfd_dev *kfd;
	85
	86	const struct kfd_device_info *device_info =
	87	lookup_device_info(pdev->device);
	88
	89	if (!device_info)
	90	return NULL;
	91
	92	kfd = kzalloc(sizeof(*kfd), GFP_KERNEL);
	93	if (!kfd)
	94	return NULL;
	95
	96	kfd->kgd = kgd;
	97	kfd->device_info = device_info;
	98	kfd->pdev = pdev;
19f6d2a6	99	kfd->init_complete = false;
4a488a7a OG	100
	101	return kfd;
	102	}
	103
b17f068a OG	104	static bool device_iommu_pasid_init(struct kfd_dev *kfd)
	105	{
	106	const u32 required_iommu_flags = AMD_IOMMU_DEVICE_FLAG_ATS_SUP \|
	107	AMD_IOMMU_DEVICE_FLAG_PRI_SUP \|
	108	AMD_IOMMU_DEVICE_FLAG_PASID_SUP;
	109
	110	struct amd_iommu_device_info iommu_info;
	111	unsigned int pasid_limit;
	112	int err;
	113
	114	err = amd_iommu_device_info(kfd->pdev, &iommu_info);
	115	if (err < 0) {
	116	dev_err(kfd_device,
	117	"error getting iommu info. is the iommu enabled?\n");
	118	return false;
	119	}
	120
	121	if ((iommu_info.flags & required_iommu_flags) != required_iommu_flags) {
	122	dev_err(kfd_device, "error required iommu flags ats(%i), pri(%i), pasid(%i)\n",
	123	(iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_ATS_SUP) != 0,
	124	(iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_PRI_SUP) != 0,
	125	(iommu_info.flags & AMD_IOMMU_DEVICE_FLAG_PASID_SUP) != 0);
	126	return false;
	127	}
	128
	129	pasid_limit = min_t(unsigned int,
	130	(unsigned int)1 << kfd->device_info->max_pasid_bits,
	131	iommu_info.max_pasids);
	132	/*
	133	* last pasid is used for kernel queues doorbells
	134	* in the future the last pasid might be used for a kernel thread.
	135	*/
	136	pasid_limit = min_t(unsigned int,
	137	pasid_limit,
	138	kfd->doorbell_process_limit - 1);
	139
	140	err = amd_iommu_init_device(kfd->pdev, pasid_limit);
	141	if (err < 0) {
	142	dev_err(kfd_device, "error initializing iommu device\n");
	143	return false;
	144	}
	145
	146	if (!kfd_set_pasid_limit(pasid_limit)) {
	147	dev_err(kfd_device, "error setting pasid limit\n");
	148	amd_iommu_free_device(kfd->pdev);
	149	return false;
	150	}
	151
	152	return true;
	153	}
	154
	155	static void iommu_pasid_shutdown_callback(struct pci_dev *pdev, int pasid)
	156	{
	157	struct kfd_dev *dev = kfd_device_by_pci_dev(pdev);
	158
	159	if (dev)
	160	kfd_unbind_process_from_device(dev, pasid);
	161	}
	162
4a488a7a OG	163	bool kgd2kfd_device_init(struct kfd_dev *kfd,
	164	const struct kgd2kfd_shared_resources *gpu_resources)
	165	{
19f6d2a6 OG	166	unsigned int size;
19f6d2a6 OG	167
4a488a7a OG	168	kfd->shared_resources = *gpu_resources;
4a488a7a OG	169
19f6d2a6 OG	170	/* calculate max size of mqds needed for queues */
	171	size = max_num_of_processes *
	172	max_num_of_queues_per_process *
	173	kfd->device_info->mqd_size_aligned;
	174
	175	/* add another 512KB for all other allocations on gart */
	176	size += 512 * 1024;
	177
	178	if (kfd2kgd->init_sa_manager(kfd->kgd, size)) {
	179	dev_err(kfd_device,
	180	"Error initializing sa manager for device (%x:%x)\n",
	181	kfd->pdev->vendor, kfd->pdev->device);
	182	goto out;
	183	}
	184
	185	kfd_doorbell_init(kfd);
	186
	187	if (kfd_topology_add_device(kfd) != 0) {
	188	dev_err(kfd_device,
	189	"Error adding device (%x:%x) to topology\n",
	190	kfd->pdev->vendor, kfd->pdev->device);
	191	goto kfd_topology_add_device_error;
	192	}
	193
b17f068a OG	194	if (!device_iommu_pasid_init(kfd)) {
	195	dev_err(kfd_device,
	196	"Error initializing iommuv2 for device (%x:%x)\n",
	197	kfd->pdev->vendor, kfd->pdev->device);
	198	goto device_iommu_pasid_error;
	199	}
	200	amd_iommu_set_invalidate_ctx_cb(kfd->pdev,
	201	iommu_pasid_shutdown_callback);
5b5c4e40	202
64c7f8cf BG	203	kfd->dqm = device_queue_manager_init(kfd);
	204	if (!kfd->dqm) {
	205	dev_err(kfd_device,
	206	"Error initializing queue manager for device (%x:%x)\n",
	207	kfd->pdev->vendor, kfd->pdev->device);
	208	goto device_queue_manager_error;
	209	}
	210
	211	if (kfd->dqm->start(kfd->dqm) != 0) {
	212	dev_err(kfd_device,
	213	"Error starting queuen manager for device (%x:%x)\n",
	214	kfd->pdev->vendor, kfd->pdev->device);
	215	goto dqm_start_error;
	216	}
	217
4a488a7a OG	218	kfd->init_complete = true;
	219	dev_info(kfd_device, "added device (%x:%x)\n", kfd->pdev->vendor,
	220	kfd->pdev->device);
	221
64c7f8cf BG	222	pr_debug("kfd: Starting kfd with the following scheduling policy %d\n",
	223	sched_policy);
	224
19f6d2a6 OG	225	goto out;
19f6d2a6 OG	226
64c7f8cf BG	227	dqm_start_error:
	228	device_queue_manager_uninit(kfd->dqm);
	229	device_queue_manager_error:
	230	amd_iommu_free_device(kfd->pdev);
b17f068a OG	231	device_iommu_pasid_error:
b17f068a OG	232	kfd_topology_remove_device(kfd);
19f6d2a6 OG	233	kfd_topology_add_device_error:
	234	kfd2kgd->fini_sa_manager(kfd->kgd);
	235	dev_err(kfd_device,
	236	"device (%x:%x) NOT added due to errors\n",
	237	kfd->pdev->vendor, kfd->pdev->device);
	238	out:
	239	return kfd->init_complete;
4a488a7a OG	240	}
	241
	242	void kgd2kfd_device_exit(struct kfd_dev *kfd)
	243	{
b17f068a	244	if (kfd->init_complete) {
64c7f8cf	245	device_queue_manager_uninit(kfd->dqm);
b17f068a OG	246	amd_iommu_free_device(kfd->pdev);
	247	kfd_topology_remove_device(kfd);
	248	}
5b5c4e40	249
4a488a7a OG	250	kfree(kfd);
	251	}
	252
	253	void kgd2kfd_suspend(struct kfd_dev *kfd)
	254	{
	255	BUG_ON(kfd == NULL);
b17f068a	256
64c7f8cf BG	257	if (kfd->init_complete) {
64c7f8cf BG	258	kfd->dqm->stop(kfd->dqm);
b17f068a	259	amd_iommu_free_device(kfd->pdev);
64c7f8cf	260	}
4a488a7a OG	261	}
	262
	263	int kgd2kfd_resume(struct kfd_dev *kfd)
	264	{
b17f068a OG	265	unsigned int pasid_limit;
	266	int err;
	267
4a488a7a OG	268	BUG_ON(kfd == NULL);
4a488a7a OG	269
b17f068a OG	270	pasid_limit = kfd_get_pasid_limit();
	271
	272	if (kfd->init_complete) {
	273	err = amd_iommu_init_device(kfd->pdev, pasid_limit);
	274	if (err < 0)
	275	return -ENXIO;
	276	amd_iommu_set_invalidate_ctx_cb(kfd->pdev,
	277	iommu_pasid_shutdown_callback);
64c7f8cf	278	kfd->dqm->start(kfd->dqm);
b17f068a OG	279	}
b17f068a OG	280
4a488a7a OG	281	return 0;
	282	}
	283
	284	void kgd2kfd_interrupt(struct kfd_dev dev, const void ih_ring_entry)
	285	{
	286	}